论文部分内容阅读
为了解决传统整数规划方法在云资源调度问题上收敛速度慢,难以适应大规模云端任务调度优化的缺陷,基于遗传算法提出了初始任务配置算法和动态任务配置算法,分别用于解决云端任务初始提交阶段和任务动态运行阶段的资源调度优化问题.在两阶段任务调度优化过程中,分别结合截止时间和资源利用率确定了有针对性的优先级队列,分别使用滑动窗口机制和在线迁移机制提升任务调度性能.通过对迭代过程和收敛速度的实验分析,本文算法能够利用遗传算法的优势解决两阶段云任务调度优化问题,并具有更快的收敛速度.
In order to solve the shortcomings of the traditional integer programming methods, such as slow convergence speed and difficulty in adapting to large-scale cloud task scheduling problem, the initial task configuration algorithm and dynamic task configuration algorithm are proposed based on genetic algorithm, which are respectively used to solve the initial submission of cloud task Phase and tasks in dynamic operation phase.On the two-phase task scheduling optimization process, the priority queues are determined according to the deadline and the resource utilization respectively, and the tasks are promoted using sliding window mechanism and online migration mechanism respectively Scheduling performance.Through the experimental analysis of the iterative process and the convergence speed, the proposed algorithm can utilize the advantages of genetic algorithm to solve the two-phase cloud task scheduling optimization problem, and has a faster convergence rate.